/MAML

Using keras and tf build MAML

Primary LanguagePython

Keras - MAML

Part 1. Introduction

As we all know, deep learning need vast data. If you don't have this condition, you can use pre-training weights. Most of data can be fitted be pre-training weights, but there all still some data that can't converge to the global lowest point. So it is exist one weights that can let all task get best result?

Yes, this is "Model-Agnostic Meta-Learning". The biggest difference between MAML and pre-training weights：Pre-training weights minimize only for original task loss. MAML can minimize all task loss with a few steps of training.

Part 2. Quick Start

  1. Pull repository.
git clone https://github.com/Runist/MAML-keras.git
  1. You need to install some dependency package.
cd MAML-keras
pip installl -r requirements.txt
  1. Download the Omiglot dataset and maml weights.
wget https://github.com/Runist/MAML-keras/releases/download/v1.0/Omniglot.tar
wget https://github.com/Runist/MAML-keras/releases/download/v1.0/maml.h5
tar -xvf Omniglot.tar
  1. Run train_mnist.py, after few minutes, you'll get mnist weight.
python train_mnist.py
235/235 [==============================] - 62s 133ms/step - loss: 0.3736 - sparse_categorical_accuracy: 0.8918
Epoch 2/3
235/235 [==============================] - 2s 9ms/step - loss: 0.0385 - sparse_categorical_accuracy: 0.9886
Epoch 3/3
235/235 [==============================] - 2s 9ms/step - loss: 0.0219 - sparse_categorical_accuracy: 0.9934
313/313 [==============================] - 27s 48ms/step - loss: 0.0373 - sparse_categorical_accuracy: 0.9882
  1. Run evaluate.py, you'll see the difference between MAML and MNIST initialization weights.
python evaluate.py
Model with mnist initialize weight train for 3 step, val loss: 1.8765, accuracy: 0.3400.
Model with mnist initialize weight train for 5 step, val loss: 1.5195, accuracy: 0.4600.
Model with maml weight train for 3 step, val loss: 0.8904, accuracy: 0.6700.
Model with maml weight train for 5 step, val loss: 0.5034, accuracy: 0.7800.

Part 3. Train your own dataset

  1. You should set same parameters in config.py. More detail you can get in my blog.
parser.add_argument('--n_way', type=int, default=10,
                    help='The number of class of every task.')
parser.add_argument('--k_shot', type=int, default=1,
                    help='The number of support set image for every task.')
parser.add_argument('--q_query', type=int, default=1,
                    help='The number of query set image for every task.')
parser.add_argument('--input_shape', type=tuple, default=(28, 28, 1),
                    help='The image shape of model input.')
  1. Start training.
python train.py --n_way=5 --k_shot=1 --q_query=1

Part 4. Paper and other implement